1
Preserving Fairness and Safety in Quantized LLMs Through Critical Weight Protection
量化大模型时公平性与安全性易受损，本文通过保护关键权重巧妙解决，让轻量化模型也能坚守底线。
arXiv:2601.12033v2 Announce Type: replace Abstract: Quantization is widely adopted to reduce the computational cost of large language models (LLMs); h…